Abstract
Epigenetic mechanisms represent potential molecular routes which could bridge the gap between genetic background and environmental risk factors contributing to the pathogenesis of pulmonary diseases. In patients with COPD, asthma and pulmonary arterial hypertension (PAH), there is emerging evidence of aberrant epigenetic marks, mainly including DNA methylation and histone modifications which directly mediate reversible modifications to the DNA without affecting the genomic sequence. Post-translational events and microRNAs can be also regulated epigenetically and potentially participate in disease pathogenesis. Thus, novel pathogenic mechanisms and putative biomarkers may be detectable in peripheral blood, sputum, nasal and buccal swabs or lung tissue. Besides, DNA methylation plays an important role during the early phases of fetal development and may be impacted by environmental exposures, ultimately influencing an individual's susceptibility to COPD, asthma and PAH later in life. With the advances in omics platforms and the application of computational biology tools, modelling the epigenetic variability in a network framework, rather than as single molecular defects, provides insights into the possible molecular pathways underlying the pathogenesis of COPD, asthma and PAH. Epigenetic modifications may have clinical applications as noninvasive biomarkers of pulmonary diseases. Moreover, combining molecular assays with network analysis of epigenomic data may aid in clarifying the multistage transition from a “pre-disease” to “disease” state, with the goal of improving primary prevention of lung diseases and its subsequent clinical management.
We describe epigenetic mechanisms known to be associated with pulmonary diseases and discuss how network analysis could improve our understanding of lung diseases.
Abstract
Network medicine approaches, supporting the integrative analysis of global epigenetic changes, clinical features and environmental factors, provide new hypotheses for pathogenesis of chronic lung diseases and lead to new interventions for prevention https://bit.ly/3kHND9m
Introduction
The molecular changes underlying the pathogenesis of many chronic pulmonary diseases are not fully understood, and potentially relevant targets for risk stratification and prevention are not well-established. In recent years, genome-wide association studies (GWAS) have identified many genomic regions that contain genetic variants significantly associated with complex, nonmalignant pulmonary diseases, including COPD, asthma and pulmonary arterial hypertension (PAH) [1–3]. However, although the development of polygenic risk scores has improved disease risk prediction, much of disease susceptibility remains unexplained by the identified variants [1–3]. Indeed, specific epigenetic marks, especially DNA methylation, histone modifications and micro-RNAs (miRNAs), can also affect the endophenotypes underlying the onset of COPD, asthma and PAH [4–7]. Furthermore, epigenetic-sensitive changes are acquired with ageing and impacted by environmental exposures or inherited by mitotic and meiotic cell division (transgenerational effects), providing a longstanding “memory” of the early life exposures experienced during fetal development that may ultimately impact an individual's susceptibility to cardiopulmonary diseases later in life [5, 8–12]. Most studies on the variability of epigenetic marks have focused on specific sites or regions of the genome. However, considering epigenetic marks in a network framework may be more informative than site-specific analyses [13–16]. A network is structured in “nodes” and “edges” which derive from applying advanced bioinformatic algorithms to omics data. This strategy has the potential to decipher the relationship between clinical data and the genes, proteins and metabolites that may play a critical role in disease pathogenesis. Remarkably, epigenetic changes modulate the interface between genomic and environmental risk factors and may drive an “accelerated” ageing of lungs [17]. Thus, investigating epigenetic changes by using network analysis in longitudinal cohorts may identify novel, noninvasive biomarkers that are useful for precision medicine applications in pulmonary disease [18–20]. Network medicine approaches combine advanced omics platforms, potent bioinformatic algorithms and clinical data [13–16]. Many network medicine applications have used the molecular interactome as a key tool to discover how the interplay between genes and the environment may differentially perturb protein–protein interactions (PPIs) [13–16]. In addition, correlation-based and gene regulatory networks have provided valuable insights into disease pathogenesis [13–16]. In this article, we summarise and discuss current evidence for the role of epigenetics in COPD, asthma and PAH, and how insights from epigenetic data are already pointing to potential biomarkers useful for prevention and personalised therapy. We also motivate the need to develop and apply network approaches to better understand the role of epigenetics and epigenetic variability in pulmonary diseases.
A focus on molecular basis of epigenetic mechanisms
Epigenetic-sensitive modifications, including DNA and RNA methylation, histone modifications and noncoding RNAs can modulate gene expression without changing DNA sequence in a spatiotemporal manner [10, 21] (figure 1). Remarkably, exposures during early development or post-natal life may perturb the individual “epi-interactome”, defined mainly by DNA methylation changes, leading to higher risk of developing complex diseases [8, 9, 22–27].
DNA methylation is a chemical reaction mediated by DNA methyltransferase (DNMT) enzymes belonging to three protein families including DNMT1, 2 or 3, which catalyse the addition of methyl groups at the 5′ carbon of cytosines into CpG dinucleotides [28]. Usually, there is an inverse relationship between promoter DNA methylation and gene expression [29, 30]
To balance methylation levels, the demethylases belonging to the 10–11 translocation (TET) family of DNA dioxygenases (TET1/2/3) remove the methyl groups from cytosine bases favouring an open chromatin state associated with active gene expression [31].
Acetylation and methylation changes are the main histone modifications which are harboured at specific amino acid positions (mainly lysine) of the N-terminal region, where they form a “code” deciphered by other chromatin remodelling complexes [32]. The global acetylation level is regulated by two classes of enzymes: histone acetyltransferases (HATs) and histone deacetylases (HDACs) which specifically increase and decrease the number of acetyl groups into histone tails, respectively. As downstream effects, HATs can favour increased gene expression, whereas HDACs are associated with gene silencing [32]. Otherwise, histone methylation is catalysed by the family of lysine methyltransferase enzymes [33] Generally, the effect of histone methylation on gene expression varies according to the specific amino acid position; for example, lysine 4 methylation is associated with gene activation [34] whereas lysine 9 methylation is associated with gene silencing [35].
miRNAs are small RNA molecules (21–22 nucleotides) acting at the post-transcriptional level where they can bind to the 3′ untranslated regions of targeted mRNAs to avoid their translation in protein products [36]. miRNA-mediated regulation is very complex because generally a single miRNA regulates hundreds of target transcripts, while multiple miRNAs can regulate the same gene [37].
Network medicine and pulmonary diseases
The pathogenesis of pulmonary diseases is triggered by complex interactions among heterogeneous molecular and environmental factors. Dissecting signalling pathways and molecular networks rather than single gene associations may allow us to clarify the aetiology and development of lung dysfunction. Here, we introduce the basic elements for understanding the framework of network-oriented analysis.
Defining molecular networks
Nodes and edges form the key components of a network-oriented analysis applied to patient-derived omics data [13–16]. In biological networks, nodes refer to the molecular players, such as genes, proteins, metabolites and noncoding RNAs, while edges represent the functional relationships between a pair of connected nodes; for example PPIs, enzymatic reactions, transcriptional and transductional regulation, post-translational protein modifications, miRNA–long noncoding RNAs (and others) [13–16]. These molecular interactions may vary over time and space in response to environmental changes, providing “dynamic networks” [13–16]. Highly studied types of molecular networks include PPI networks, which are based on the human molecular interactome, and gene regulatory networks, which are based on transcription factor–DNA and protein–ligand binding [13–16]. In addition, co-expression networks are constructed to find modules of highly correlated genes based on their expression profiles (or co-methylation profiles) [13–16]. Biological networks are organised through quantifiable rules. For example, most biological networks have a topology that is “scale-free”, i.e. the node distribution follows a power law [38]. From a functional point of view, network hubs (nodes with many connections) represent the roots of connectivity among essential genes that can lead to embryonic lethality if mutated, whereas peripheral nodes (less highly connected nodes located far from the centre of the network) often represent genes harbouring the genetic variants that are responsible for phenotypic heterogeneity or that contribute the most to complex disease risk [13–16]. Thus, a network analysis can provide insights into the behaviour of complex biological systems and suggest putative useful clinical biomarkers or drug targets by identifying, for example, dysregulated hubs in the PPI network.
Network analysis in transition from health to disease
The “disease module hypothesis” states that the nodes and edges relevant to disease pathogenesis are more likely to be located in the same region, or “subnetwork”, rather than being randomly scattered throughout the molecular interactome [13–16]. Many advanced network-oriented algorithms have been applied to patient-derived data to define the functional and mechanistic pathways underlying pulmonary diseases [4, 39–50]. These have generally included two main approaches: 1) identifying novel candidate genes localised in lung disease modules and 2) measuring the distance between the module in which the putative candidate gene is present and a known module for a disease [13–16].
Network analyses of lung diseases are currently often performed by analysing biological samples isolated from end-stage patients and control subjects. This approach only provides associative evidence between disease phenotype and molecular biomarkers and thus does not offer a temporal modelling of the dynamic molecular networks responsible for key transition events from health to disease. However, chronic respiratory disease pathogenesis is probably a multistage process moving from a healthy state toward a pre-disease state (reversible phase) which can transition to a disease state (irreversible or reversible phase). Thus, there is a need to perform analyses on the evolution of the relevant biological networks across healthy, pre-disease state (at risk, asymptomatic) and disease state subjects in order to pick up early network signals that initiate the critical transition to disease; these may point to useful biomarkers for primary prevention [51].
The proposed hypothesis of “fetal origins” of pulmonary diseases: epigenetics for primary prevention?
The hypothesis supporting fetal origins for human disease arises from a convincing relationship between environmental exposures, lower birthweight and higher risk of cardiovascular dysfunction later in life [52]. Epigenetic-sensitive changes could bridge the mechanistic gap between many environmental and genetic risk factors providing a long-lasting memory of early detrimental exposures in COPD and asthma [18–20, 44], as well as in PAH [53]. DNA methylation plays a key role during fetal development; thus, large prospective birth cohorts have investigated DNA methylation as a surrogate for association between maternal [18–20] and/or paternal [19] exposures and risk of asthma and COPD in offspring. The largest meta-analysis of cord blood-based studies, including 1688 children from five cohorts revealed that specific differentially methylated regions (DMRs) were associated with childhood lung function, childhood asthma and COPD in adulthood. Among the top annotated genes, DMRs located in the homeobox A5 (HOXA5) gene promoter, involved in morphogenesis of fetal lung, correlated to childhood/adolescent forced expiratory volume in 1 s (FEV1) and COPD development [18]. Moreover, hypermethylation of the promoter of the SMAD3 (small mother against decapentaplegic 3) gene in neonatal life predicted the risk of asthma in childhood in those individuals born to asthmatic mothers [20]. Interestingly, a birth cohort of 1629 newborns revealed that the presence of a father with a smoking history was associated with persistent hypermethylation of immune master players, including the genes LIM domain only 2 (LMO2) and interleukin 10 (IL-10) and development of childhood asthma [19]. This evidence supports the hypothesis that maternal/paternal transgenerational effects can perturb biological processes in offspring via DNA methylation modifications. Thus, fetal epigenetic programming may influence COPD and asthma risk during adulthood. Methylome profiling at birth might eventually be a useful clinical tool to identify subjects with higher risk of pulmonary diseases and improve opportunities for primary prevention (figure 2). PAH has less support for epigenetic programming [5]. Although still debated in humans, fetal programming may impact epigenetic-sensitive pathways underlying the inheritance of pulmonary diseases and understanding these pathways may reveal new opportunities for primary and primordial prevention.
Differential epigenetic biomarkers for prevention of pulmonary diseases in post-natal life
Here, we summarise results about putative epigenetic-sensitive biomarkers useful to measure the individual risk for pulmonary disease onset and progression (tables 1–3) [54–74]. We use these studies as exemplars of progress in the field toward the identification of epigenetic marks for further rigorous study in network analyses and functional validation. In addition, we summarise a list of studies from https://clinicaltrials.gov and https://eudract.ema.europa.eu/ in supplementary table S1 to highlight the increasing interest in clinical epigenetics [75–78].
COPD
COPD is characterised by progressive obstruction of airflow associated with emphysematous destruction of lung parenchyma, small airway fibrosis and destruction, and/or hypersecretion of mucus with chronic bronchitis [79]. Smoking is the most powerful risk factor for COPD development; as an exposure, smoking has been investigated in humans from in utero to old age and demonstrates extensive epigenetic effects and is a model of the impact of environmental exposure on the human genome. The epigenetic impact of smoking persists years after smoking cessation [80, 81]. However, some other inhaled noxious particles and gases, such as biomass combustion products, are also established COPD risk factors [79].
DNA methylation
There are many studies investigating the association between DNA methylation, COPD and lung function, mostly based on the analysis of circulating blood cells [54, 55] and less on lung tissue [56] providing putative useful biomarkers. An epigenome-wide association study (EWAS) of pulmonary function and COPD (274 COPD cases versus 2919 controls) based on peripheral blood samples identified 28 differentially methylated CpG sites associated with pulmonary function levels and/or COPD, mapping to genes involved in alternative splicing, JAK-STAT signalling and axon guidance; further research into DNA methylation associations may improve prediction of COPD risk [54]. In addition, a publicly available dataset of COPD case–control lung tissue was retrieved to investigate the functional impact of DNA methylation changes on gene expression, suggesting an enrichment for JAK-STAT signalling genes including the suppressor of cytokine signalling 3 (SOCS3) [54]. This warrants further functional validation experiments in large cohorts to establish a possible causal relationship between methylation changes and COPD. A longitudinal EWAS identified targeted meta-stable DNA methylation changes at CpG dinucleotides in peripheral blood strongly associated with lung function decline and ageing in individuals from two independent cohorts (discovery set n=633; replication set n=868) [55]. Meta-stable methylation loci annotated to the aryl-hydrocarbon receptor repressor (AHRR) gene were associated with reduced FEV1, forced expiratory flow at 25–75% of forced vital capacity and ageing, whereas the immediate early response 3 (IER3) gene was associated with reduced FEV1, independent of ageing, suggesting a potential biomarker to predict lung function decline in the general population [55]. By combining genome-wide DNA methylation analysis of lung tissue samples from 114 COPD patients (all former smokers) versus 46 healthy controls with previous GWAS results, Morrow et al. [56] reported an enrichment for CpG differential methylation in shelves and shores annotated to several transcription factors, such as the forkhead box k1 (FOXK1) and FOXP2 transcription factors, as well as asthma-related genes, including the M1 muscarinic acetylcholine receptor (CHRM1) gene. However, there has been limited consistency between studies of the associations of lung function or COPD with differentially methylated sites, and one limiting factor may be related to the cellular heterogeneity of lung tissue and whole blood.
Histone modifications
Few studies have focused on histone modifications to clarify the molecular basis of lung inflammation in COPD pathogenesis. Sundar et al. [82] demonstrated that cigarette smoke exposure triggered I κB kinase α (IKKα)-mediated phosphorylation (activation) of the mitogen- and stress-activated kinase 1 (MSK1) with consequent phospho-acetylation of histone H3 (Ser10/Lys9) and acetylation of histone H4 (Lys12) harboured at NF-κB-dependent promoters in H292, BEAS-2B and SAEC cells, leading to transcriptional activation of genes probably involved in COPD pathogenesis. Higher expression of β-defensin 1 (DEFB1) was associated with higher histone deacetylase 1 mRNA levels and correlated with poor lung function parameters in bronchopulmonary specimens of patients with COPD (n = 34) versus healthy controls (n = 10) [64].
Micro-RNAs
miR-145 negatively regulated the release of pro-inflammatory cytokines, such as IL-6 and C-X-C motif chemokine ligand 8 (CXCL8), by targeting SMAD3 in airway smooth muscle cells isolated from nine COPD patients versus 18 healthy controls [69]. Interestingly, further administration of miR-145 mimics reduced IL-6 and CXCL8 expression in COPD airway smooth muscle cells to levels comparable to controls, suggesting a novel putative drug therapy [69]. By using a microarray approach, Savarimuthu Francis et al. [70] found a significant downregulation of miR-34c, miR-34b, miR-149, miR-133a and miR-133b in lung tissue from 29 COPD patients with moderate emphysema versus mild emphysema undergoing resection for lung cancer. Among them, miR-34c showed the largest degree of association and modulated expression of its putative target mRNAs, such as the serpin family e member 1 (SERPINE1) gene.
Asthma
Asthma is a characterised by chronic inflammation of the airways leading to respiratory symptoms including wheezing, shortness of breath, chest tightness, cough and variable airflow limitation, which vary over time and in intensity [83]. Asthma global prevalence ranges from 1% to 21% in adults and up to 20% of children aged 6–7 years [84].
DNA methylation
Recently, an EWAS conducted on genomic DNA from nasal epithelial cells of 547 children reported a strong correlation between elevated biomarkers of allergic disease, including fractional exhaled nitric oxide and total IgE, and differentially methylated regions that were annotated to genes involved in structure and function of epithelial cells, oxidative stress and mucin production enzymes in asthmatics versus never-asthmatics [57]. A consistent signal arose from hypomethylation status at the eosinophil peroxidase (EPX) gene promoter which has been implicated in driving type 2 T-helper (Th2) cell and eosinophilic responses; however, there were no statistically significant associations between DNA methylation and lung function [57]. Furthermore, a DNA methylome analysis in peripheral blood mononuclear cells (PBMCs) from three independent cohorts (n=394 subjects) including mild-to-moderate asthmatic patients found that hypomethylation both of the IL-12 subunit β (IL12B) and precortistatin (CORT) genes correlated with decreased hospitalisation and oral corticosteroid use, respectively, suggesting useful prognostic biomarkers [58]. A large family-based EWAS on circulating eosinophils isolated from 95 European pedigrees demonstrated significant associations between serum IgE and methylation levels at 36 promoter regions annotated to genes coding for eosinophil products and phospholipid inflammatory mediators, specific transcription factors and mitochondrial proteins [59]. Hypomethylation of genes encoding eosinophil products was highly representative of asthmatics with high IgE versus asthmatics with low IgE levels and controls, suggesting potential risk stratification biomarkers [59]. More recently, an integrated methylome-single cell transcriptome analysis was performed by collecting nasal epithelial brush samples from 455 children (aged 16 years) with asthma, rhinitis, and asthma and rhinitis (AsRh) [60]. In these data, replicable differentially methylation sites were associated with the rhinitis and AsRh phenotypes (but not with asthma) and suggested useful disease predictors which were enriched for immune pathways, such as microglia pathogen phagocytosis pathway, DAP12 interactions, adaptive immune system, IL-2 signalling pathway and T-cell receptor signalling pathway [60]. Another multi-omics approach has investigated the impact of differential methylation on gene expression in bronchial biopsies from 26 persistent asthmatic patients and 39 asthma patients in remission versus 70 healthy controls [61]. In this study, four CpG-sites and 42 DMRs were found in persistent asthma versus the remission group. In particular, the most significant CpG site and DMR were hypomethylated in remission versus persistent asthma and associated with lower expression of the atypical chemokine receptor 2 (ACKR2) in subjects with remission [61]. The ACKR2 gene mediates resolution of inflammation, suggesting a potential molecular mechanism involved in the remission of asthma.
Histone modifications
Some studies clarified the mechanisms of histone modulation controlling cytokine production and its role in the onset/severity of allergic diseases. A genome-wide mapping of histone modifications in CD4+ memory, Th1 and Th2 cells from 12 asthmatic patients versus 12 healthy subjects revealed a differential enrichment of histone H3 lysine 4 di-methylated (H3K4me2) in Th2 enhancers associated with asthma susceptibility [66]. Moreover, a significant enrichment for binding sites of transcription factors involved in T-cell differentiation, such as GATA3, TBX21 and RUNX3, as well as an over-representation of genes involved in chemokine and Toll-like receptor signalling pathways, were observed [66]. In addition, hyperacetylation of histone H4 at the glucocorticoid response element upstream of the dual specificity phosphatase 1 gene (DUSP1), encoding for mitogen-activated protein kinase (MAPK) phosphatase 1 (MKP-1), was found in monocytes treated with dexamethasone obtained from eight steroid-sensitive and 11 steroid-resistant asthmatics patients. Additionally, vitamin D pre-incubation increased the dexamethasone-induced H4 acetylation, providing anti-inflammatory effects [65].
Micro-RNAs
A panel of circulating serum miRNAs, including miR-146b, miR-206 and miR-720, which are involved in NF-κB and GSK3/AKT pathways, might improve the accuracy of asthma exacerbation risk prediction in a paediatric asthma cohort (n=153) [71]. Indeed, the integrated miR-clinical score model showed high predictive power in discriminating exacerbation versus no exacerbation (area under the receiver operating characteristic curve 0.81) [71]. Maes et al. [72] investigated the possible association between miRNA expression in sputum with the inflammatory cell profile from 17 patients with mild-to-moderate asthma, nine patients with severe asthma and 10 healthy individuals. After validation in a second independent cohort, upregulation of miR-629-3p, miR-223-3p and miR-142-3p were significantly associated with severe neutrophilic asthma suggesting putative useful biomarkers to predict disease severity [72].
PAH
PAH is a relatively rare cardiopulmonary disease characterised by irreversible vascular remodelling of pulmonary arterial walls leading to progressive vasoconstriction, inflammation, and thrombosis which may culminate in right ventricle failure [5, 84]. Vascular cells in PAH patients share some cancer hallmarks, such as hyperproliferation, resistance to apoptosis, migration and metabolic switch toward glycolysis (Warburg effect) by affecting epigenetic-sensitive pathways [5].
DNA methylation
DNA methylation may play a role in PAH [5], but clinical data are limited. The first DNA methylome analysis of cultured pulmonary arterial endothelial cells (PAECs) isolated from 11 subjects with idiopathic PAH and 10 subjects with heritable PAH (HPAH) versus 18 healthy controls revealed that the most significant hypermethylated region mapped to the ATP-binding cassette 1 (ABCA1) promoter, suggesting that cholesterol metabolism is closely associated with PAH [62]. Moreover, the targeted bisulfite sequencing of DNA extracted from PBMCs revealed that the promoter region of the bone morphogenetic protein receptor type 2 (BMPR2) gene, which represents a relevant driver in PAH pathogenesis, was hypermethylated in the wild-type allele of 12 HPAH patients versus controls (unaffected relatives) suggesting a putative regulatory mechanism underlying the low penetrance of disease [63]. The next step is to perform integrative analysis which correlates methylome signatures in inflammatory circulating cells or disease-relevant tissue with quantitative gene expression to better understand PAH pathogenesis.
Histone modifications
Histone acetylation imbalance is relevant in vascular remodelling of PAH patients, and the impaired gene expression and enzymatic activity of the extracellular superoxide dismutase (SOD3 or EC-SOD), a major vascular antioxidant enzyme, was related to activation of HDAC3 in lung tissue from end-stage PAH patients at time of lung transplantation versus lungs from failed donors [67]. The activity of the transcription factor myocyte enhancer factor 2 (MEF2), a key regulator of lung homeostasis, was significantly reduced in PAECs isolated from 10 PAH patients versus seven controls leading to lower levels of miR-424 and 503, connexins 37, and 40 and Krüppel-like factors 2 and 4 [68]. This detrimental effect was correlated to an excess of nuclear HDAC4 and HDAC5 activity. Moreover, treatment of PAECs with a small interfering RNA able to selectively target HDAC4 and HDAC5 led to reactivation of the MEF2 and expression of its gene targets, with associated decreased cell migration and proliferation, suggesting a therapeutic value in PAH [68].
Micro-RNAs
Recently, a multiple epigenetic-sensitive regulatory axis, integrating phosphatidylinositol 3-kinase (PI3K)-DNMT1-miR-1281-HDAC4 signalling, has been implicated in proliferation and migration of pulmonary arterial smooth muscle cells (PASMCs) under platelet-derived growth factor stimulation [73]. In particular, low levels of miR-1281 were identified in 19 paediatric patients with PAH versus 13 healthy subjects, suggesting a possible diagnostic use [73]. Moreover, a strong correlation between low circulating miR-34a-3p levels and overexpression of mitochondrial dynamics protein of 49 and 51 kDa (MiD49 and MiD51, respectively) was found to play a role in promoting the Warburg effect in PASMCs from PAH patients versus controls, suggesting further useful biomarkers for disease treatment [74].
Network approaches to modelling epigenetic variability in primary prevention of pulmonary diseases
Currently, the application of network analysis to epigenetic data is quite limited in pulmonary diseases; there has been greater focus on genetic variation affecting molecular mechanisms leading to disease phenotypes [40–43]. However, recent integrative omics studies have started to use network-oriented analyses to model interactions among biological, environmental and clinical features to develop novel hypotheses regarding lung disease mechanisms and to prioritise relationships for further validation. For example, network-oriented analyses have been applied to epigenetic data in fetal lung tissue exposed to in utero smoke and COPD [44], as well as in heterogeneous patient-derived cells or tissues in asthma [46, 47] and PAH [48, 49]. Weighted gene correlation network analysis (WGCNA) is a systems biology approach for identifying correlation patterns among the biological molecules measured in high-throughput experimental assays. WGCNA has been applied to DNA methylation data from fetal and adult lung tissue, revealing preserved network modules associated with fetal exposure to in utero smoke, COPD and lower adult lung function [44]. These modules were significantly enriched for genes involved in embryonic organ development and specific inflammation-related and ageing pathways, including Hippo, PI3K/AKT, Wnt, MAPK and transforming growth factor-β signalling, potentially supportive of the fetal origins of COPD [44]. Separately, a WGCNA analysis of leukocyte DNA methylation from 362 African American individuals with and without COPD identified comethylation modules associated with COPD that were enriched for genes related to inflammatory pathways, lung development and immune response, suggesting possible candidate genes contributing to racial differences in COPD susceptibility and severity [45]. In addition, by using linear regression and Pearson correlation analysis miRNA–mRNA pair co-expression networks were built, revealing that 20 miRNAs and 539 mRNAs were associated with chronic mucus hypersecretion (CMH) in bronchial biopsies of 63 COPD patients (from mild CMH to moderate/severe CMH) versus healthy controls belonging to the GLUCOLD study (NCT00158847) [46]. The gene set enrichment analysis and PPI-related STRING tool also predicted that upregulated miR-134-5p, miR-146a-5p and the let-7 family and their potential target genes, including the KRAS proto-oncogene, GTPase (KRAS) and endothelin 1 (EDN1), may be key drivers of CMH in COPD [46]. Additionally, a molecular-bioinformatic approach applying WGCNA to DNA methylation and RNA sequencing data in primary airway epithelial cells from 74 asthmatic and 41 healthy subjects revealed that DNA methylation plays a relevant role in mediating the effects of individual genetic background on asthma risk and clinical outcome [47]. In addition, regulatory network analyses have provided important insights into pulmonary disease. For example, Reyes-Palomares et al. [48] built a multi-omic regulatory network of PAECs isolated from explanted lungs of 10 end-stage PAH patients and nine donor control individuals. Analysis of this network suggested that remodelling of active enhancers (marked by H3K27ac), rather than poised enhancers (marked by H3K4me1) and promoters (marked by H3K4me3), may prime PAH-PAECs towards the endothelial–mesenchymal transition without affecting gene expression [48]. Moreover, blood-based downregulation of miR-140-5p levels was found in treatment-naïve idiopathic PAH patients (n=4, female) compared to controls (n=4, female) [49]. Interestingly, network analysis predicted the SMAD-specific E3 ubiquitin protein ligase 1 (SMURF1) gene as a target for miR-140-5p target and a hub of bone morphogenetic protein (BMP) signalling [49]. This study also demonstrated that SMURF1 expression was increased in whole blood and remodelled pulmonary vasculature of PAH patients versus controls and that treatment with a miR-140-5p mimic may restore BMP signalling, suggesting a novel strategy to counteract onset and development of disease [49]. Furthermore, Parikh et al. [50] used miRNA relationships as filters to build a disease network module for PAH. Validation in PAECs revealed a key regulatory role for miR-21 upregulation in disease pathogenesis by impacting BMPR2 gene expression.
Novel perspectives
The “early life” stage, ranging from preconception (influenced by in utero exposures) through early childhood, is a critical temporal window during which detrimental insults may shape epigenetic signatures affecting fetal developmental and health later in life [8, 9, 18–20, 26, 44, 53]. Capturing tissue-specific epigenetic modifications by using network approaches might increase our knowledge about the fetal origins of diseases. The LifeCycle Project (https://lifecycle-project.eu/) is a European network of birth cohorts focusing on the role of early epigenetic-sensitive biomarkers, mainly DNA methylation, related to socioeconomic, migration, urban environment and lifestyle risk factors underlying pulmonary (and other) diseases. This consortium may be the starting point for investigating the possible impact of maternal and paternal lifestyle and nutritional habits on epigenetic-sensitive changes associated with pulmonary diseases by using a network-oriented strategy. This would have important implications for personalised diet and lifestyle interventions that may facilitate primary and primordial prevention of disease. Although the hypothesis of the fetal origins of diseases has gained international attention in recent years due to its potential in preventing complex diseases, there are still no definite strategies for implementation. Beyond obvious challenges from scientific and economic points of view, a critical barrier is the failure to communicate a key message to policymakers [85]. When elaborating on the possible implications of network medicine approaches for early life interventions for pulmonary diseases, two major questions should be clarified to the research community. First, how do we go from molecular networks to validated biomarker signatures to either predict disease risk or stratify patients for “individualised” therapies? Second, what is specifically needed to move from descriptive studies to actually implementing epigenetic biomarkers in clinical practice? Only future prospective randomised clinical trials will tell us whether specific early-life network-oriented epigenetic signatures may have a clinical utility in pulmonary disease management.
Conclusions
Epigenetic-sensitive modifications have been associated with chronic lung diseases, suggesting a pathogenic role for epigenetic marks in disease onset and progression. Remarkably, DNA methylation may be inherited across several generations (transgenerational effect). This highlights the clinical potential for early-life methylome analysis in cord blood cells as a way to quantify the risk for disease onset, as observed for COPD and asthma, although the data are less compelling for PAH (figure 2). However, evidence for the role of cell-specific epigenetic marks in lung disease pathogenesis remains associative in the current clinical settings. In addition, it is still unknown whether most of the DMRs that are found in cord blood cells really reflect methylation in nasal cells or lung tissue. Therefore, these results cannot be used as predictors of disease in individuals, but are only useful from an aetiological perspective. In addition, DNA methylation patterns and expression of genes vary depending on the developmental stage [86] and are not be stable in the same individuals from birth to adulthood. Furthermore, DNA methylation marks are widely associated with sex [87–89] and age (epigenetic clock) [90–92] implying potential sex- and age-specific epigenomic responses to pre-natal exposures, which may serve as biomarkers of disease susceptibility. Thus, further integrative clinical studies should be performed to assess whether blood cells mirror the aberrant DNA methylation features and downstream modifications of gene expression patterns in diseased lungs. By exploring molecular pathways, network-oriented algorithms have the potential to clarify the mechanistic links between exposures, epigenetic (and genetic) variability, and lung disease pathogenesis, offering novel opportunities for primary prevention and precision medicine. The involvement of specific pathways underlying the immune response seem to be an overlapping feature of COPD, asthma and PAH. Thus, it would be useful to further explore aberrant epigenetic features by multi-omics platforms and network-oriented analysis in selected circulating immune cells to detect molecular pathways which might be good signals to identify novel drug targets or for repurposing of “old drugs” (drug repurposing). However, the incomplete knowledge of the human interactome, the inaccuracy of gene ontology annotation, the lack of standard methods for validating network predictions, as well as the absence of randomised trials testing the clinical utility of network-based biomarkers, still limit the incorporation of network medicine to a bench approach in pulmonary precision medicine [5, 13, 15].
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary table S1 ERJ-03406-2020.SUPPLEMENT
Graphical abstract ERJ-03406-2020.ABSTRACT
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-03406-2020.Shareable
Footnotes
This article has supplementary material available from erj.ersjournals.com
Author contributions: G. Benincasa, D.L. DeMeo, E.K. Silverman and C. Napoli contributed to the study concept and design. G. Benincasa, D.L. DeMeo, K. Glass, E.K. Silverman and C. Napoli contributed to acquisition, analysis, or interpretation of data. G. Benincasa, D.L. DeMeo and K. Glass drafted the manuscript. G. Benincasa, D.L. DeMeo and C. Napoli did critical revision of the manuscript for important intellectual content. G. Benincasa prepared the figures.
Conflict of interest: G. Benincasa has nothing to disclose.
Conflict of interest: D.L. DeMeo has nothing to disclose.
Conflict of interest: K. Glass has nothing to disclose.
Conflict of interest: E.K. Silverman has nothing to disclose.
Conflict of interest: C. Napoli has nothing to disclose.
Support statement: This manuscript was funded by National Research Competitive Grant (PRIN supporting grant no. F8ZB89) from Ministry for University and Research, Italy (to C. Napoli). G. Benincasa is a PhD Student in Translational Medicine supported from University of Campania “Luigi Vanvitelli,” Naples, Italy. This study was also funded by US National Institutes of Health grants U01 HL089856, P01 HL114501, P01 HL132825, R01 HL147148, R01 HL133135, R01 HL137927 and K25 HL133599. E.K. Silverman reports institutional grant funding from GSK and Bayer. D.L. DeMeo reports institutional grant funding from Bayer and a research grant from the Alpha-1 Foundation. Funding information for this article has been deposited with the Crossref Funder Registry.
- Received September 6, 2020.
- Accepted November 10, 2020.
- Copyright ©ERS 2021